Loki API - How many queries per second can it handle?


I’m building a platform where I may want to expose my loki logs to the customers using the Loki API. I was wondering if this would be okay in terms of potential load that we might get, is Loki designed to be able to handle high number of queries per second? And does the QPS scale horizontally with the number of loki nodes?


I am not running a huge Loki cluster myself, but so far I am pretty happy with its scalability. What’s a high number you have in mind?

If you are looking to run a scalable Loki cluster, I would recommend:

  1. Deploy in either microservice mode (individual components) or simple scalable mode (some co-existing components).
  2. As part of deploying in either of the two modes above, you’d naturally be separating read and write traffic, which allows you to scale read or write path separately.
  3. Use Query Frontend along with querier for distributed querying.